Five AI Labs Agree to Pre-Release US Government Reviews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Google, Microsoft, and xAI joined OpenAI and Anthropic in agreeing to grant the U.S. Commerce Department's CAISI early access to evaluate their frontier AI models prior to public release
- CAISI has now signed formal agreements with Google DeepMind, Microsoft, and xAI regarding national security testing of frontier AI, following earlier deals with OpenAI and Anthropic
- OpenAI and Anthropic renegotiated their existing arrangements with Washington to align with the new pre-release screening framework for AI model evaluations
- The U.S. government will conduct security reviews of new AI models from top labs through a voluntary process that lacks legal mandate but represents the closest thing to current federal AI oversight
- Andrew Curran noted that the new pre-release screening agreements with CAISI across all five major AI labs may be tied to an upcoming AI executive order and policy memo expected the same day
Why it matters: Five leading AI labs are now aligned under a voluntary U.S. government review process for frontier models, creating a de facto oversight standard without legislation. This shifts leverage to federal agencies despite the lack of legal authority, raising stakes for future regulatory mandates.
